Simulations for Sociology: An Agent-Based Model of Peer-Influence testing the Protective Effects of Social Capital on Adolescent Substance-Use

Abstract

Adolescent substance-use remains a significant societal concern, with implications for individual well-being and community welfare. Drawing on Icelandic sociological research and core sociological theories, this thesis seeks to computationally model the phenomenon of deviant peer-influence and adolescent substance-use by testing different experimental conditions of peer-smoking and social capital among agents. This research uses agent-based simulations to examine the complexities of adolescent substance-use, focusing on the protective impact of social capital and the influential role of deviant peers. I hypothesize that agents who are more highly connected to their communities with higher social capital should experience less substance-use and receive less deviant influence from their peers. To conduct this research, I simulated varying friend group compositions to observe the outcomes of differing group dynamics based on social attributes of the agents. Agent friend groups, comprised of six agents each, were simulated through 20 school days with different combinations of high and low levels of social capital and high, medium, and low levels of peer-group smoking. Social capital was represented as a cumulative value based on an agent’s involvement with sports, time spent with their parents, and their commitment to school, and are based upon the domains of intervention underlying the Icelandic Prevention Model. Levels of deviance were represented using smoking behavior among agents, who all have a propensity to smoke but will only do so if their urge to smoke surpasses a certain threshold. In these simulations, an agent’s urge to smoke was influenced by their peers and social capital value alike throughout the 20 school days. The results of the simulations reveal that agents with greater levels of social capital were least impacted by the influence of peer-group smoking behavior, and that group dynamics heavily influenced group substance-use. These findings are in line with what was hypothesized based on the empirical research from the Icelandic literature, namely that social capital would serve as a protective factor against the influence of peer substance-use. This thesis contributes to the development of a sub-model of peer-influence within the Social Insight framework, an ontology-driven agent-based model of risky adolescent behavior. Additionally, these findings contribute to the ongoing theoretical discussion on the complex and dynamic nature of adolescent substance-use and the ways in which we may employ novel techniques to study such phenomenon.